内容来源：https://www.theverge.com/news/788741/microsoft-copilot-labs-ai-portraits-experiment-human-avatar

内容总结：

微软公司近日为其人工智能助手Copilot推出了一项名为“人像对话”的实验性功能，为AI助手赋予了可实时互动的动态虚拟形象。该功能目前面向美国、英国和加拿大的成年用户开放测试，提供40种风格化人物肖像作为对话载体，在语音交流过程中可呈现自然的面部表情与口型变化。

微软人工智能首席执行官穆斯塔法·苏莱曼在社交平台X上表示，该功能旨在回应用户对“可视化对话界面”的需求。通过将选定肖像与语音系统结合，使AI语音交互获得更接近人际对话的体验。这是继今年7月推出“Copilot外观”功能后，微软在提升AI交互亲和力方面的又一次尝试。与之前的卡通化设计不同，新功能采用基于真人形象设计的风格化肖像。

技术层面，该实验运用了微软研究院开发的VASA-1人工智能系统，仅需单张图像即可实时生成自然的面部活动，无需复杂三维建模。值得注意的是，微软在功能部署上保持审慎态度：所有肖像均采用非写实风格设计，并设置了使用时长限制和明确的AI交互标识。公司强调，该功能在提供视觉多样性的同时，仍将保持Copilot原有的智能水平与安全防护机制。

此举正值多家科技企业布局具身AI交互领域之际。此前竞争对手X公司已为其Grok聊天机器人推出包含动漫风格虚拟伴侣在内的3D形象功能。微软采取相对稳健的推进策略，或源于行业对AI交互安全性的持续关注。

中文翻译：

微软正尝试为Copilot人工智能助手添加动画面部形象，使其更具亲和力。这项名为“肖像”的实验功能目前在美国、英国和加拿大上线，提供40款风格化虚拟人像，可在实时语音对话中展现自然表情。

微软AI首席执行官穆斯塔法·苏莱曼在X平台宣布该功能时表示，肖像模式的开发源于部分Copilot用户反馈“更希望在语音交互时能看到对话对象的面孔”。用户可选择特定肖像并匹配语音，旨在让与聊天机器人的语音对话比纯文字交流更自然。

这是微软将Copilot塑造为更具吸引力交互对象的又一尝试。此前七月份该公司曾推出类似功能“Copilot外观”，但新功能的显著区别在于采用真人形象而非动画卡通造型。微软在Copilot的Discord服务器中说明：“本实验运用微软研究院开发的VASA-1先进人工智能技术实现实时可视化AI对话，仅需单张图像即可生成自然面部表情、头部动作与唇形同步，无需复杂3D建模。”

此举正值竞争对手X公司为其Grok聊天机器人推出3D虚拟形象之际，其中包含具备NSFW模式的动漫风格AI女友。微软对肖像功能采取谨慎推广策略，可能源于对用户与聊天机器人间有害互动的安全考量——此类问题正使Character.AI等平台接受调查。

微软在Copilot博客中强调：“所有肖像均采用刻意风格化设计而非写实风格，在保持您信赖的Copilot智能与安全特性同时提供视觉多样性。”目前该功能仅限18岁以上“特定用户群体”使用，设有单日及单次会话时长限制，并配备“明确的人工智能交互提示标识”。

英文来源：

Microsoft is trying to make Copilot more approachable by giving the AI assistant an animated face to talk with. The experimental “Portraits” feature in Copilot Labs is currently available in the US, UK, and Canada, and provides 40 stylized human avatars that respond with natural expressions during real-time voice conversations.
Microsoft is giving Copilot AI faces you can chat with
Copilot Portraits provides 40 stylized human avatars that make expressions during voice conversations.
Copilot Portraits provides 40 stylized human avatars that make expressions during voice conversations.
Announcing the new feature on X, Microsoft AI CEO Mustafa Suleyman says Portraits was developed in response to some Copilot users saying “they’d feel more comfortable talking to a face when using voice.” The feature allows users to select a portrait and pair it with a voice, aiming to make verbal conversations with the chatbot feel more natural compared to text-based interactions.
This is Microsoft’s latest attempt to develop Copilot into a character that’s more appealing to communicate with, having previously announced a similar feature called “Copilot Appearances” in July. The notable difference between the two is that Portraits is based on human appearances instead of an animated cartoon blob.
”This experiment leverages VASA-1, an advanced AI technology developed by Microsoft Research to create real-time, visual AI conversations,” Microsoft announced on the Copilot Discord server. “The innovation allows us to generate natural facial expressions, head movements, and lip-sync from just a single image — no complex 3D modeling required.”
The Portraits experiment follows rival AI provider X introducing 3D avatars for its Grok chatbot, including an anime-inspired AI girlfriend companion with an NSFW mode. Microsoft is taking things slow with the rollout of Portraits, likely due to safety concerns around harmful interactions between users and rival chatbots, like those under investigation from Character.AI.
“Each portrait is intentionally stylized and not photorealistic, offering you visual variety while offering the same trusted Copilot intelligence and safety features you rely on,” Microsoft said on the Copilot blog. Availability is restricted to a “limited set of users” aged 18 or older, according to Microsoft, with daily and session-based time limits in place alongside “clear indicators that you are interacting with AI technology.”